[https://github.com/liaopeiyuan/artbench](ArtBench) samples encoded to float16 SDXL latents via [Ollin VAE](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix).

Dataset created using [this script](https://github.com/Birch-san/sdxl-diffusion-decoder/blob/main/script/make_sdxl_latent_dataset.py).

Didn't bother saving mean & logvar, because variance is low enough it's not worth the doubling of filesize to retain.  
Sampled from diagonal gaussian distribution, saved the resulting latents.  
Also kept the original image.

Schema/usage:

```python
from typing import TypedDict, Iterator
from webdataset import WebDataset
Sample = TypedDict('Sample', {
  '__key__': str,
  '__url__': str,
  'cls.txt': bytes, # UTF-8 encoded class id from 0 to 9 inclusive
  'img.png': bytes, # PIL image, serialized. 256*256px
  'latent.pth': bytes, # FloatTensor, serialized. 32*32 latents
})

it: Iterator[Sample] = WebDataset('train/{00000..00004}.tar')

for sample in it:
  pass
```

The data sources of ArtBench-10 is released under a Fair Use license, as requested by WikiArt, Ukiyo-e.org database and The Surrealism Website.  
For more information, see https://www.wikiart.org/en/terms-of-use, https://ukiyo-e.org/about and https://surrealism.website/ 

train: 50000 samples  
test: 10000 samples

```python
# test/avg/val.pt (mean):
[-0.11362826824188232, -0.7059057950973511, 0.4819808006286621, 2.2327630519866943]
# test/avg/sq.pt:
[52.59075927734375, 30.115631103515625, 44.977020263671875, 30.228885650634766]
# std
# (sq - val**2)**.5
[7.251058578491211, 5.442180633544922, 6.689148902893066, 5.024306297302246]
# 1/std
[0.1379109025001526, 0.18374986946582794, 0.14949584007263184, 0.19903245568275452]

# train/avg/val.pt (mean):
[-0.1536690890789032, -0.7142514586448669, 0.4706766605377197, 2.24863600730896]
# train/avg/sq.pt:
[51.99677276611328, 30.184646606445312, 44.909732818603516, 30.234216690063477]
# std
# (sq - val**2)**.5
[7.2092413902282715, 5.447429656982422, 6.68492317199707, 5.017753601074219]
# 1/std
[0.1387108564376831, 0.18357281386852264, 0.14959034323692322, 0.1992923617362976]
```